1. Field of the Invention
This invention relates generally to the field of equalizers. More particularly, the invention provides a transmit amplitude independent adaptive equalizer that is capable of compensating for transmission losses in an input signal when the transmit signal amplitude is unknown. The invention is particularly well suited for use in digital communication components, such as receivers, equalizers, high-speed backplanes, Printed Circuit Board Trace equalizers, automatic gain control devices, and other types of digital communication components.
2. Description of the Related Art
The use of an equalizer to compensate for loss resulting from the non-idealities of a transmission medium is known. FIG. 1 is a block diagram showing an equalizer 12 implemented in a typical digital communications system 10 in which an input signal 14 is transmitted through a transmission medium 16. Typical transmission media 16 used for transmission of digital signals over relatively short distances include, for example, printed circuit board (PCB) traces and coaxial cables. These, and other known transmission media, typically cause significant frequency dependant losses in digital signals being transmitted over the media and consequently distort the digital data, often resulting in pulse spreading and interference between neighboring pulses (known as intersymbol interference). In addition, the input signal 14 is further corrupted during transmission by noise 18 induced by the transmission medium 16. The equalizer 12 regenerates the transmitted signal 20 by providing gain to compensate for the frequency dependant losses caused by the transmission medium 16 (up to some maximum length) while preferably minimizing the effect of noise 18. This function is typically achieved by applying a transfer function to the received signal 20 that approximates the inverse of the transmission losses.
FIG. 2 is a graph 30 showing the loss (in dB) incurred in the transmission medium 16, plotted as a function of both the length (l) of the medium 16 and the frequency (f) of the signal. Generally, the loss over a transmission medium (such as a coaxial cable or PCB trace) may be approximated in the frequency domain by the following equation:L(f)=e−l(ks√{square root over (jf)}+kd|f|);where f is the frequency, l is the length of the transmission medium, j=√{square root over (−1)}, ks is the skin effect loss constant of the transmission medium, and kd is the dielectric loss constant of the transmission medium. The value of L(f) is plotted in FIG. 2 for transmission media of two different lengths: Length 1 (shorter) and Length 2 (longer). As the length (l) of the transmission medium increases, the loss increases. In addition, as the frequency (f) increases, the loss increases.
To counteract the transmission loss shown in FIG. 2, an equalizer 12 should have a frequency characteristic that is the inverse of the loss function of the transmission medium. The inverse loss function may be approximated as follows:             1              L        ⁢                  (          f          )                      =                  G        ⁢                  (          f          )                    =              1        +                  KH          ⁢                      (            f            )                                ;where K is a control variable that is proportional to the length (l) of the transmission medium. The value of K typically varies from zero to unity (or some other constant) as the transmission medium approaches its maximum length.
FIG. 3 is a graph 40 showing the inverse loss function G(f), plotted in dB on the same axes as the loss function L(f). As shown in this figure, the inverse loss function G(f) provides a frequency dependant gain equivalent to the loss L(f) incurred in the transmission medium. The characteristics of the inverse loss function G(f) are explained in more detail in U.S. patent application Ser. No. 09/055,515 (hereinafter referred to as “the '515 application”) which is owned by the Assignee of the present application, and which is hereby incorporated into the present application by reference.
FIG. 4 is a block diagram of an equalizer core 50 that implements the inverse loss function G(f). The equalizer core 50 includes a transfer function block 52 (H(f)), a multiplier 58, and an adder 56. This circuit 50 applies variable gain to an input signal 57 by applying the transfer function H(f) in order to generate a resultant signal and then by multiplying the resultant signal from the transfer function block 52 by a gain control signal 58 (K). The gain control signal 58 (K) preferably controls the amount of gain applied by the transfer function H(f) by multiplying the output of the transfer function block 52 by a factor typically varying from zero (0) to unity (1) depending upon the length (l) of the transmission medium 16. For instance, when the transmission medium 16 is at a maximum length, the transfer function H(f) is generally multiplied by unity (1) to provide the maximum gain. The output of the multiplier is then summed with the input signal 57 by the adder 56 in order to produce an equalized output signal 59 corresponding to the inverse loss function (1+KH(f)). An exemplary circuit for implementing the transfer function block 52 is described in the above-referenced '515 application.
FIG. 5 is a block diagram of an alternative equalizer core 60 that implements a bandwidth-limited inverse loss function. In this circuit 60, a low-pass filter 62 is added to the equalizer core 50 shown in FIG. 4 to reduce noise encountered in the transmission medium 16. This alternative implementation 60 reduces the amplification of high frequency noise, and thus increases the signal-to-noise ratio (SNR) of the equalized output signal 64. A graphical representation 70 of the bandwidth-limited inverse loss function 72, plotted on the same axes as the loss function L(f) is shown in FIG. 6.
FIG. 7 is a block diagram showing a multiple-stage equalizer core 80 having three equalizer stages 82, 84 and 86, each of which implements the inverse loss function G(f). The three cascaded equalizer stages 82, 84 and 86 are preferably the same as the equalizer core 50 shown in FIG. 4. Alternatively, the multiple-stage equalizer core 80 could include a plurality of bandwidth-limited stages as shown in FIG. 5, or other types of cores. In any case, each equalizer stage 82, 84 and 86 includes a gain control signal (K1, K2 or K3) that is used to control the gain implemented by the transfer function H(f) in proportion to the length of the transmission medium 16. The advantages of utilizing a multiple-stage equalizer core are explained in detail in the '515 application.
Operationally, each stage 82, 84 and 86 in the multiple-stage equalizer core 80 is configured to equalize signals transmitted over transmission media up to a percentage of the total maximum transmission medium length. For instance, if the multiple-stage equalizer core 80 is capable of equalizing losses incurred in a printed circuit board (“PCB”) trace of up to 30 inches, then each core stage 82, 84, and 86 will typically be configured to equalize losses in PCB traces of up to 10 inches. The stages 82, 84 and 86 are then cascaded such that they operate sequentially to equalize PCB traces of up to 30 inches.
FIG. 8 is a graph 90 showing how the gain control signals K1, K2 and K3 in the multiple-stage equalizer core 80 are varied according to the length of the transmission medium. The value K, shown along the x-axis in FIG. 8, represents the percentage of the transfer function H(f) that needs to be applied to an input signal in order to supply the gain necessary to equalize a transmission medium of a given length. As the transmission medium length increases, the gain necessary to equalize the transmission losses in the medium also increases. FIG. 8 shows that the gain control signals K1, K2 and K3 cause gain to be supplied sequentially by the equalizer stages 82, 84 and 86. For instance, if each equalizer stage 82, 84 and 86 is capable of providing the necessary gain to equalize 10 inches of a PCB trace, then the gain control signal K1 would typically control the gain necessary for PCB traces from 0 to 10 inches, the combined gain control signals K1 (at unity) and K2 would typically control the gain necessary for PCB traces from 10 to 20 inches, and the combined gain control signals K1 (at unity), K2 (at unity) and K3 would typically provide the gain for PCB traces from 20 to 30 inches. For example, if the PCB trace were 15 inches in length and each equalizer stage 82, 84 and 86 can equalize 10 inches, then K1 would be at its maximum value (unity), K2 would be at the value necessary to cause the second equalizer stage 84 to equalize a 5 inch transmission medium, and K3 would be zero.
FIG. 9 is a block diagram showing an exemplary equalizer system 100 such as described in the referenced '515 application. This equalizer system 100 includes an equalizer core 102, a slicer 104, an automatic gain control circuit (AGC) 106, a transmitter 108, and a transmission medium 110. The equalizer core 102 may be either a single-stage core as shown in FIG. 4 or 5 or a multiple-stage core as shown in FIG. 7, and operates, as described above, to compensate for the losses incurred in the transmission medium 110. The output 112 of the equalizer core 102 is coupled to the slicer 104, which converts the output signal 112 from the core 102 to a digital output signal 114 having a known swing (A) that approximates the swing (B) of the data sent from the transmitter 108. Since the swing (B) of the transmitted data is known and reproduced as the swing (A) of the digital output signal 114 from the slicer 104, the difference in energy between the equalizer core output signal 112 and the digital output signal 114 approximates the energy lost in the transmission medium 110, which is proportional to its length. The AGC 106 compares the energy of the equalizer core output signal 112 with the energy of the digital output signal 114 from the slicer 104 to generate the gain control signal K.
The AGC 106 includes a core-side band-pass filter 116, a core-side envelope detector 118, a slicer-side band-pass filter 120, a slicer-side envelope detector 122, an adder 124, and a sequencer 126. Operationally, the AGC 106 filters the core and digital outputs 112 and 114 to mid-band frequencies using the band-pass filters 116 and 120. The advantage of filtering the core and digital outputs 112 and 114 to their mid-band frequencies is explained in detail in the '515 Application. Following this filtering function, the AGC 106 then detects the signal energy of the two band-limited signals with the envelope detectors 118 and 122. Finally, it determines the difference between the two signal energies with the adder 124, which provides the gain control signal K. If the equalizer core 102 is single-stage, then the gain control signal K is typically coupled directly to the core 102 to control the variable gain as described above. If, however, the equalizer core 102 is of the multiple-stage type, then the sequencer 126 is used to convert the gain control signal K from the adder 124 into a plurality of multiple-stage gain control signals Ki, such as K1, K2 and K3 described above with reference to FIGS. 7 and 8. In either case, the gain control signal(s) K (or Ki) enable the equalizer core 102 to equalize the core output signal 112 by forcing it to the same energy level as the digital output signal 114 from the slicer 104. A further description of the AGC 106 is provided in the above referenced '515 application.
One skilled in the art will appreciate that the signal swing (B) at the transmitter 108 must be known a priori and accurately replicated by the slicer 104 if the equalizer system 100 shown in FIG. 9 is to achieve optimal performance. Any significant difference between the signal swing (B) at the transmitter 108 and the signal swing (A) of the digital output signal 114 will directly result in a gain (equalization) error. For example, an increase in the swing (B) of the transmitted signal will force the AGC loop 106 to settle at a lower gain than necessary to compensate for the transmission loss (under-equalization). Similarly, a decrease in the swing (B) of the transmitted signal will result in over-equalization. Even if the swing (B) of the transmitted signal were tightly controlled, similar equalization errors may be caused by mismatch in the digital output swing (A) generated by the slicer 104. Such mismatch errors may be caused, for example, by variations in temperature, power supply voltages, or manufacturing processes.